Academic Journals Database
Disseminating quality controlled scientific knowledge

29 A Novel Approach for Identify Small and Capital Handwritten Letter

ADD TO MY LIST
 
Author(s): Ms. Ekta Tiwari | Dr. Maneesh Shreevastava

Journal: International Journal of Advanced Computer Research
ISSN 2249-7277

Volume: 2;
Issue: 2;
Start page: 29;
Date: 2012;
VIEW PDF   PDF DOWNLOAD PDF   Download PDF Original page

Keywords: OCR | Features | Support Vector Machine (SVM) | Artificial Neural Networks | Handwritten Character Recognition | Stroke | Printed Characters.

ABSTRACT
A handwritten character is represented as asequence of strokes whose features are extractedand classified. Although the off-line and on-linecharacter recognition techniques have differentapproaches, they share a lot of common problemsand solutions. The printed documents available inthe form of books, papers, magazines, etc. arescanned using standard scanners which produce animage of the scanned document. The preprocessedimage is segmented using an algorithm whichdecomposes the scanned text into paragraphs usingspecial space detection technique and then theparagraphs into lines using vertical histograms, andlines into words using horizontal histograms, andwords into character image glyphs using horizontalhistograms. Each image glyph is comprised of24x24 pixels. Thus a database of character imageglyphs is created out of the segmentation phase. Thevarious features that are considered forclassification are the character height, characterwidth, the number of horizontal lines (long andshort, image centroid and special dots. we proposedextracted features were passed to a Support VectorMachine (SVM) where the characters are classifiedby Supervised Learning Algorithm. These classesare mapped onto for recognition. Then the text isreconstructed using fonts.
Affiliate Program      Why do you need a reservation system?